Masters of suspicion: A Bayesian decision model of motivated political reasoning

Abstract

Motivated reasoning occurs when judgements subserve motives that go beyond accuracy seeking. Substantial evidence indicates that motivated political reasoning is ubiquitous. This is hard to reconcile with computational theories (following Marr's terminology, theories describing the fundamental principles underlying a cognitive process) like Bayesian inference, because these rely on accuracy maximization. Hence, motivated political reasoning is often interpreted as violating computational principles. Here we propose a different view by offering a computational account of motivated political reasoning which relies on the notion of Bayesian decision (instead of Bayesian inference). The key idea is that utility maximization, and not accuracy maximization, drives political thinking. This implies that agents will tend to endorse judgements that serve their instrumental goals even when evidence in support is poor (though agents will still believe their judgements are the most accurate). In this framework, motivated political reasoning is not interpreted as violating computational principles, although its nature is now conceived as pragmatic (i.e., serving instrumental goals) rather than epistemic (i.e., seeking understanding). The paper presents a mathematical description of the theory and shows how this can help interpreting important phenomena in political psychology such contextual priming, stereotyping, and displaced aggression.